You are on page 1of 5

Empirical Models

Douglas Plaza Guingla, Ph.D.


November 21, 2019

1 Introduction
Normally, for the purpose of tuning a control loop is not worth to spend
much time in developing a model. Instead Input and Output process data is
record by performing a plant test. The aim of the plant test is to obtain a
model which represents the dynamics of the system.
There are some plant tests in the literature. Here, the step plant test is
explained.

2 First order plus dead time (FOPDT)


A first order plus dead time system is represented by the following transfer
function:
kp e−θs
Y (s) = U (s) (1)
τp s + 1
where kp is the process gain, θ is the time delay and τp is the time constant.
The output response of a FOPDT model is given by:
− t−θ
 
y(t) = kp ∆u 1 − e τp (2)

the steady-state gain kp of the process is obtained from equation 2 when


t → ∞.
∆Y
kp = (3)
∆U

1
2.1 One point method for estimating the time param-
eters
Let’s evaluate equation 2 when time is t = τp + θ, the expression is the
following:

y(τp + θ) = kp ∆U (1 − e−1 )
y(τp + θ) = 0.632kp ∆U

according to the last expression, the time t63.2 = τp + θ is the time that the
response takes to get to 63.2% of its steady-state magnitude kp ∆U . Thus,
the value of the time constant τp can be identified according to the 63.2%
criteria. Figure 1 shows the one point method to identify the time constant
of the process.
y(t)

63.2% of ∆Y
∆Y

time
t63.2%

Figure 1: Step response to obtain the model

2.2 Two point method to identify the time constant


Two point method is based on the one point method, additionally, a second
point when the output response reaches 28.3 % of its magnitude ∆Y = kp ∆U .
The expressions for obtaining the time constant and the time delay are given

2
as follows:
τp = 1.5(t63.2% − t28.3% ) (4)
θ = t63.2 − τp (5)
Figure 2 shows the two point method, which is a graphically method.
y(t)

63.2% of ∆Y
∆Y

28.3% of ∆Y

time
t28.3% t63.2%

Figure 2: two point method

There is a limitation in the application of the step test to identify a


FOPDT model. The limitation is related to the time spent in the acquisition
of the gain of the process (steady-state conditions). In the industrial field,
processes exhibits large time constant.

3 Integrator plus dead time


The transfer function representation of an integrating process with time delay
is given by:
ke−θs
Y (s) = U (s) (6)
s
the corresponding response in the time domain is:
y(t) = k∆U (t − θ) (7)
thus k∆U represents the slope of the curve as it can be seen in figure 3.

3
y(t)

slope = k∆U
slope
k= ∆U

time

Figure 3: integrating process response

Table 1: Ziegler-Nichols Open-Loop tuning parameters FOPDT


Controller type kc τI τD
τp
Proportional kp θ
- -
0.9τp
PI kp θ
3.3θ -
1.2τp
PID kp θ
2θ 0.5θ

4 Controller tuning based on FOPDT/IPDT


parameters
In this section the gains of the controllers (P, PI and PID) are presented:

4
Table 2: Ziegler-Nichols Open-Loop tuning parameters FOPDT
Controller type kc τI τD
τp
Proportional kp θ
- -
0.9τp
PI kp θ
3.3θ -
1.2τp
PID kp θ
2θ 0.5θ

Table 3: Ziegler-Nichols Open-Loop tuning parameters IPDT


Controller type kc τI τD
1
Proportional kθ
- -
0.9
PI kθ
3.3θ -
1.2
PID kθ
2θ 0.5θ

You might also like